lux-ai-challenge / lux-design-s2 Goto Github PK

New API removes can_x functions as submitting an action queue incurs power as well so it could be confusing.

Moreover, default random agent should check move_cost is not null and move_cost + action_queue_cost <= unit.power

I'm not sure if this is a lux AI bug or a pettingzoo, but when I attempt to run the validate code to make sure everything is installed correctly, I'm met with an error.

I have tried uninstalling and reinstalling pettingzoo with no success

Code:
luxai2022 my_bot/main.py my_other_bot/main.py -v 2 -o replay.json

Error:

Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Scripts\luxai2022.exe\__main__.py", line 4, in <module>
  File "C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Lib\site-packages\luxai_runner\cli.py", line 3, in <module>
    from luxai_runner.bot import Bot
  File "C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Lib\site-packages\luxai_runner\bot.py", line 9, in <module>
    from luxai_runner.logger import Logger
  File "C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Lib\site-packages\luxai_runner\logger.py", line 3, in <module>
    from luxai2022.globals import TERM_COLORS
  File "C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Lib\site-packages\luxai2022\__init__.py", line 1, in <module>
    from .env import LuxAI2022
  File "C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Lib\site-packages\luxai2022\env.py", line 7, in <module>
    from pettingzoo import ParallelEnv
ImportError: cannot import name 'ParallelEnv' from 'pettingzoo' (C:\Users\lgntr\AppData\Local\Programs\Python\Python311\Lib\site-packages\pettingzoo\__init__.py)

If I run the getting started steps from the manual I get the following error when trying to validate my installation:

Both using (https://github.com/Lux-AI-Challenge/Lux-Design-2022):
luxai2022 my_bot/main.py my_other_bot/main.py -v 2 -o replay.json

and (https://github.com/Lux-AI-Challenge/Lux-Design-2022/tree/main/kits/python)
luxai2022 main.py main.py --out=replay.json

Version info:
python 3.7.13
luxai2022 1.0.4

The errors are:
Task exception was never retrieved
future: <Task finished coro=<BotProcess.start() done, defined at DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\process.py:33> exception=NotImplementedError()>
Traceback (most recent call last):
File "DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\process.py", line 47, in start
limit=1024 * 128
File "DIR\Anaconda3\envs\lux\lib\asyncio\subprocess.py", line 217, in create_subprocess_exec
stderr=stderr, **kwds)
File "DIR\Anaconda3\envs\lux\lib\asyncio\base_events.py", line 1544, in subprocess_exec
bufsize, **kwargs)
File "DIR\Anaconda3\envs\lux\lib\asyncio\base_events.py", line 462, in _make_subprocess_transport
raise NotImplementedError
NotImplementedError
Task exception was never retrieved
future: <Task finished coro=<BotProcess.start() done, defined at DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\process.py:33> exception=NotImplementedError()>
Traceback (most recent call last):
File "DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\process.py", line 47, in start
limit=1024 * 128
File "DIR\Anaconda3\envs\lux\lib\asyncio\subprocess.py", line 217, in create_subprocess_exec
stderr=stderr, **kwds)
File "DIR\Anaconda3\envs\lux\lib\asyncio\base_events.py", line 1544, in subprocess_exec
bufsize, **kwargs)
File "DIR\Anaconda3\envs\lux\lib\asyncio\base_events.py", line 462, in _make_subprocess_transport
raise NotImplementedError
NotImplementedError
Traceback (most recent call last):
File "DIR\Anaconda3\envs\lux\lib\runpy.py", line 193, in _run_module_as_main
"main", mod_spec)
File "DIR\Anaconda3\envs\lux\lib\runpy.py", line 85, in run_code
exec(code, run_globals)
File "DIR\Anaconda3\envs\lux\Scripts\luxai2022.exe_main.py", line 7, in
File "DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\cli.py", line 83, in main
asyncio.run(eps.run())
File "DIR\Anaconda3\envs\lux\lib\asyncio\runners.py", line 43, in run
return loop.run_until_complete(main)
File "DIR\Anaconda3\envs\lux\lib\asyncio\base_events.py", line 587, in run_until_complete
return future.result()
File "DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\episode.py", line 109, in run
resolved_actions = await asyncio.gather(*action_coros)
File "DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\bot.py", line 54, in step
action, stderr = await asyncio.wait_for(self.proc.write(f"{data}\n"), timeout=self.remainingOverageTime + self.time_per_step)
File "DIR\Anaconda3\envs\lux\lib\asyncio\tasks.py", line 442, in wait_for
return fut.result()
File "DIR\Anaconda3\envs\lux\lib\site-packages\luxai_runner\process.py", line 72, in write
self._agent_process.stdin.write(msg.encode())
AttributeError: 'BotProcess' object has no attribute '_agent_process'

• encourage more map movement
• prevents solutions that can simply mine adjacent / super close tiles

Currently this is silently ignored and the CLi tool automatically turns it into a action queue for some reason.

• Basic 2D grid visuals, fast and renders new states quickly
• Replay scrubber (speed up speed down, play and stop)
• Button to upload a replay file
• Setup distribution. (build files should include an html file that can then request all the built js files)

It is possible to build factories that overlap.

I think this should lead to an error instead?

Extend luxai_runner tool with a --playback functionality. Playback option can include logging error logs/warnings/info (e.g. moving without enough power, crashing into opposition) into the replay to be viewed on the replay viewer.

Also add a batch --playback option by setting --parallel and passing a directory of replays instead of a single one.

• Lichen growth simulation as a gif
• What map looks like
• Different units
• Weather effects
• Collision gifs

Still extremely complicated, and there's ambiguity with regards to building factories

One user ran

tar -cvf submission.tar.gz *

to tar.gz their submission and submit it. The issue here is that kaggle competition requires the -z option, so you must run

tar -cvzf submission.tar.gz *

An issue posed by some competitors in season 1 that is more likely to be a more problematic season 2 as the action space has more dimensions and a lot of them are useless depending on certain values due to the human-friendly parameterization of all possible actions.

Python kit currently does simple strategy

• Bid nothing and randomly spawn factories with 100 metal and water
• Build heavy unit as soon as possible
• Each unit moves to closest ice tile and mines until it has 100 ice
• Units move to closest factory tile, transfer it to factory, repeat.

Improve the pygame visualizer (useful for quick inspection of map changes / progress)

Right now it doesn't render the timestep...
Does not render team colors correctly, units correctly...